Ultimate Tic-Tac-Toe Reinforcement Learning

A RL project using dual-head ResNet and MCTS (AlphaZero-like).
Includes arena for matches between models and GUI to play against your model.

Setup

uv sync

How to run

Some of these commands may require you to already have a model file to use.
Each command has several configurable arguments which you can check by appending --help.

Main training loop

uv run main.py

Arena

uv run arena.py

Play against model with GUI

uv run play.py